Security

@antv npm Attack Hits CI/CD Secrets and Tokens

3 min read

Summary

Microsoft has disclosed an active supply chain attack involving compromised @antv npm packages that used malicious preinstall scripts to steal credentials from GitHub Actions and other CI/CD environments. The campaign matters because it spread through popular downstream dependencies, putting developer pipelines, cloud secrets, and software supply chains at risk.

Need help with Security?Talk to an Expert

Introduction

Microsoft has uncovered a significant npm supply chain attack affecting the @antv package ecosystem. Because these packages are widely used directly and transitively, the compromise can reach far beyond a single project and into CI/CD pipelines, cloud environments, and downstream software builds.

For IT and security teams, this is a reminder that package ecosystem compromises can quickly become credential theft and cloud access incidents.

What happened

A threat actor compromised an @antv maintainer account and published malicious package versions. The malicious code executed during npm install through a preinstall hook and targeted GitHub Actions runners on Linux.

Key technical findings

  • Malicious @antv packages propagated through dependency chains into widely used libraries such as echarts-for-react
  • The payload was heavily obfuscated and designed to evade analysis
  • It executed silently during package installation
  • It focused on CI/CD environments, especially GitHub Actions
  • It installed Bun if needed and launched a second-stage payload

Credentials and secrets targeted

Microsoft says the malware attempted to steal credentials from multiple platforms, including:

  • GitHub tokens and repository secrets
  • AWS credentials and Secrets Manager data
  • HashiCorp Vault tokens
  • npm tokens
  • Kubernetes service account secrets
  • 1Password CLI data

The payload also scraped GitHub Actions runner process memory to extract secrets directly, potentially bypassing normal secret masking protections.

Why this matters for administrators

This is more than a developer issue. If affected packages were installed in enterprise build systems, attackers may have gained access to:

  • Source code repositories
  • CI/CD secrets and pipelines
  • Cloud workloads and infrastructure
  • Package publishing credentials

Microsoft also notes the malware attempted privilege escalation, dual-channel exfiltration, and even SLSA provenance forgery, which undermines trust in software supply chain attestations.

What GitHub did

GitHub responded by:

  • Removing 640 malicious packages
  • Invalidating 61,274 npm granular access tokens with write permissions and 2FA bypass
  • Publishing advisories and issuing Dependabot and npm audit alerts

Administrators should act quickly:

  • Review dependency trees for direct or transitive use of affected @antv packages
  • Identify systems that installed or built these packages during the exposure window
  • Rotate GitHub, npm, AWS, Vault, Kubernetes, and other potentially exposed credentials
  • Audit GitHub Actions workflows, runner logs, and repository changes for suspicious activity
  • Check for unauthorized repos, commits, secrets access, or package publication events
  • Pin known-good package versions and strengthen dependency controls

Bottom line

The @antv compromise shows how a single npm maintainer account takeover can cascade into widespread CI/CD credential theft. Security and DevOps teams should treat any exposure as a potential pipeline and cloud credential incident, not just a package hygiene issue.

Need help with Security?

Our experts can help you implement and optimize your Microsoft solutions.

Talk to an Expert

Stay updated on Microsoft technologies

npmsupply chain securityGitHub ActionsCI/CDcredential theft

Related Posts

Security

Microsoft Gaming Security: Key Risks and Defenses

Microsoft’s latest Deputy CISO post explains why securing gaming requires a different approach than traditional enterprise IT. The company outlines the distinct risks across gaming platforms, studios, and shared central teams, and highlights how Entra ID, Purview, Defender for Cloud, and Sentinel help balance security with player experience and developer agility.

Security

Microsoft RAMPART and Clarity Open-Sourced

Microsoft has open-sourced RAMPART and Clarity, two tools aimed at improving safety in agentic AI development. RAMPART brings repeatable adversarial and regression testing into CI pipelines, while Clarity helps teams challenge design assumptions early before code is written.

Security

Fox Tempest Malware Signing Service Disrupted

Microsoft has disrupted Fox Tempest, a malware-signing-as-a-service operation that helped cybercriminals make ransomware and other malware appear legitimately signed. The takedown matters because the group abused Microsoft Artifact Signing, created more than 1,000 fraudulent certificates, and enabled attacks that could bypass security controls more easily.

Security

Storm-2949 Cloud Breach: Entra ID to Azure Attack

Microsoft detailed how Storm-2949 turned a socially engineered Microsoft Entra ID compromise into broad data theft across Microsoft 365 and Azure. The case highlights how identity attacks can escalate quickly through legitimate cloud management features, making stronger MFA controls, monitoring, and cross-platform detections critical for defenders.

Security

Microsoft Security for SMBs in an AI-Powered World

Microsoft is urging small and medium businesses to treat cybersecurity as a core business risk as AI makes phishing, malware, and identity attacks faster and more effective. The company highlights Microsoft 365 Business Premium and integrated security controls as a practical way for growing businesses to protect users, devices, email, and cloud apps without adding major complexity.

Security

Autonomous AI Agents: Microsoft Defense-in-Depth

Microsoft outlines a defense-in-depth approach for securing autonomous AI agents as they move from assisting users to taking actions across systems. The guidance emphasizes that the application layer—not just the model—is the most important control point for limiting permissions, enforcing human review, and reducing blast radius in production.